DESCRIPTION (Applicant's Abstract): HIV/AIDS prevention for Latinas is urgently needed. The AIDS pandemic has continued to increase among women, especially women of color. In Chicago, AIDS is the leading cause of death among Latinas age 25-44 years. No research has documented the effectiveness of a culturally-tailored intervention for adult Latinas. The specific aim of this study is to evaluate the effectiveness of a culturally-specific intervention, Project SEPA, in increasing HIV prevention behaviors for inner city Mexican and Puerto-Rican women. The content and learning strategies of Project SEPA are based on the integrated social learning model of AIDS prevention shown to be the most effective in increasing Aids prevention behaviors, modified to take into account the special needs of Latinas related to gender inequality and cultural values and practices. The culturally-relevant issues of unequal power, based on the cultural value of Machismo/Marianismo conflict management and prevention of violence, are specifically addressed. Community-based peer groups are especially compatible with Latino culture. The conceptual basis for the intervention aims to facilitate later replication of the intervention in different communities. The intervention will be using a randomized pre-test/post-test comparison design and follow-up assessments immediately and 3 and 6 months after completion of the intervention. Approximately 768 sexually-active Mexican and Puerto- Rican women aged 18 to 40 years will be recruited into the study. Because individual women (in the intervention condition) are nested within groups, the design requires the use of multi-level (or hierarchical analysis. Any initial group differences can be adjusted for using pre-test measures as co-variates in the regression models. The major aim of the study is having an intervention that is culturally specific to the needs of Latinas. Additional aims includes having a randomized control group and the longitudinal design with data collection points at pre-test, post-test, 3 month follow-up and 6 month follow-up. The design also controls for historical effects and language by randomizing within waves and within language groups.